Work
Where I've worked, what I owned, and what shipped.
Product and Growth Intern
Amuse Labs, Internship
Summary
At Amuse Labs, I worked on retention and growth for India Mini, a D2C daily crossword built on PuzzleMe, the company's B2B puzzle engine. Using funnel analysis and Microsoft Clarity heatmaps, I designed five UX experiments around streaks and login behaviour, then implemented and measured two in production. I generalised that analysis into a reusable retention pipeline with an agent-ready spec, which we also ran on a Times of India puzzle launch. I also built a crawl pipeline to find and score US publisher leads, benchmarking competitors like LinkedIn and Duolingo along the way. Most of all, I learned to turn drop-off data into a story about user behaviour, and to work with UX and engineering teams at industry standard.
Highlights
- Lifted D7 streak retention 40%+ by shipping 2 production UX experiments, after funnel analysis diagnosed login frequency as retention's lever
- Flagged 35% recoverable lapsing users through a survival-curve segmentation model, redirecting the feature roadmap toward win-back mechanics
- Surfaced 190+ premium US publisher targets via a 5,000+ site crawl pipeline scoring traffic and competitor gaps, redefining sales strategy
- Built a reusable insight engine spanning 830K+ plays and 120K+ users by generalising the retention study, later powering the new launch of India's largest daily
- Standardised retention metrics for India's largest daily by documenting definitions, edge cases and anomalies in an agent-ready spec, across 830K+ plays
Product and Strategy Intern
TheAdmitCo., Internship
Summary
At Admit Co, an early-stage venture founded by ISB alumni that runs GMAT tutoring and portfolio counselling for MBA aspirants, I joined as a product and strategy intern. Student tracking ran over WhatsApp and scattered sheets and docs, so I sat with the founders and a few students to find the pain points and built small solutions for each. I automated recurring manual work with Notion and Google Apps Script, rebuilt the workspace in Notion (80+ documents, 15+ templates), and connected it to an LMS I built, which 30+ students were onboarded onto. I also redesigned the website. On the growth side, I ran outreach through Reddit and other social media, and college WhatsApp channels, which lifted enrolment ~30% in 2 weeks and added ~₹1.2L in revenue. It was end-to-end ownership: find the problem, build the fix, then put it in front of users.
Highlights
- Drove ~30% enrolment growth in 2 weeks, adding ~₹1.2L in program revenue through targeted outreach and college-network promotions
- Built a self-developed LMS that tracks each student's full journey, and onboarded 30+ students onto it in place of informal WhatsApp and sheet-based tracking
- Migrated 80+ documents into a centralised Notion workspace with 15+ templates, linked directly to the LMS
- Automated recurring daily tasks with Notion automations and Google Apps Script, and redesigned the website
AI Systems & Full Stack Mentor
CreatED, Internship
Summary
At CreatED, a startup that helps students build real-world projects with industry mentors, I worked as a software mentor across a 6-month engagement. I took 10+ problem statements and turned them into buildable specs, owning scoping, architecture and technical direction across 500+ hours. I guided two beginner students through full prototypes under a strict zero-cost constraint. One was PurePath, a routing app for Delhi cyclists that makes live AQI part of the "best route". The other was a RAG-based AI tutor on Postgres and pgvector, where benchmarking Gemini against GPT on 600+ questions got us to ~90% accuracy. Mentoring taught me to design for what a beginner can understand and maintain, and to leave out complexity the prototype didn't need.
Highlights
- Turned 10+ problem statements into buildable specs, owning technical direction across 500+ hours of AI analysis and cloud architecture
- Software mentor for student development projects, owning scoping, technical direction and delivery across a 6-month engagement
- Designed a zero-cost, RAG-powered AI tutor by benchmarking Gemini and GPT on 600+ questions, reaching ~90% accuracy with Gemini
- Built PurePath, a routing app that ranks cycling routes by live air quality as well as distance and time, with no database, to keep the build lightweight
AI Analysis & Systems Engineering Intern
Aether AI, Internship
Summary
At Aether AI, an emerging healthtech startup, I worked as an AI Analysis and Full Stack Development intern on an AI medical assistant that reads handwritten prescriptions. I led the research and pipeline work: image preprocessing, benchmarking multiple OCR models word by word, fine-tuning Microsoft TrOCR on a custom dataset, and a fuzzy-match layer over a medicine database for reliable extraction. I also designed and built most of the client portal frontend, wiring the pipeline into a fine-tuned MedLlama model alongside the GPT API. Along the way I learned to choose models by studying where they fail on messy real-world data, and to ship quickly with co-interns and full-time staff in a startup setting.
Highlights
- Lifted prescription accuracy to 80% from 65% by analyzing failure patterns across 5+ modeling approaches, then presented findings to founders and investors
- Closed a data gap for Indian handwriting recognition by compiling and labeling 300+ real prescription images, since no public benchmark existed
- Turned a two-month sprint into the startup's flagship AI medical product, shipping a 5-interface portal from problem framing to live demo



